Re: [Xen-devel] Update Seabios on xen-unstable

[Fabio Fantoni <fantonifabio@xxxxxxxxxx>]

Il 02/01/2013 13:10, Ian Campbell ha scritto:
> On Thu, 2012-12-20 at 12:34 +0000, Fabio Fantoni wrote:
> > I saw good news about qemu-xen on xen-unstable updates to latest stable
> > version (1.3), I already start test it and report bugs found.
> > I saw no seabios updates for now on xen-unstable, there is 1.7.1
> > upstream version since few months that include all vgabios.
> > Is possible to update seabios on xen-unstable please?
> > I have tried it time ago for probably resolution of qxl vga problem
> > without result.
> > It probably needs some particular settings or modification for correct
> > integration in xen about vgabios that I not know of.
> > Thanks for any reply.
> This is on my todo list, however it is made a bit tricky because SeaBIOS
> is quite picky about the compiler which is used (specifically the newest
> SeaBIOS won't build on a Debian Stable system, which affects both me and
> the Xen.org test system)
>
> The solution will probably be to ship a binary and only rebuild when
> explicitly asked to. I haven't quite got round to it yet though.
>
> Ian.
Thanks for your reply.
Seabios 1.7.1 is now on Debian sid repository, I tested it with Wheezy 
and xen-unstable and it is working.

Qemu still crashes if qxl is active and used but I don't think it is a 
seabios problem.
I've been looking for the problem in xen but I didn't find a solution at 
the moment. I can tell that seabios is called by hvmloader that sets a 
few things like acpi and e820 mapping before calling it. Could it be 
that hvmloader does something that interferes with seabios, like 
imposing restrictions?
In practice if I activate qxl but I use it only as stdvga I don't 
experience any problem except low video performance, but if I use it 
fully qemu crashes when spice start using qxl features, logging this error:
(/usr/sbin/xl:4658): Spice-CRITICAL **: red_memslots.c:123:get_virt: 
slot_id 194 too big, addr=c2c2c2c2c2c2c2c2
The only other possibility that comes to mind is that not enough 
videoram is successifully set. I thought this problem was already solved 
some months ago with xen and qemu upstream patches. Could there be other 
problems preventing the assignment and use 128 mb of videoram?
Thanks for any reply.

Attachment: smime.p7s
Description: Firma crittografica S/MIME

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel